Who it is forAn agent operator whose host hits context limits when an upstream tool returns a very large result.
The problemA large tool response floods the context window, and re-running the upstream call to see more of it may be costly or change the result.
The outcomeThe host reads the exact stored snapshot in bounded fragments or by cursor, and the upstream tool is not invoked again.
What you need first
- A reviewed checkout of the lennney/mcp-slim-guard repository (0.1.1-alpha.2 is not yet published to npm)
- npm to run ci, build, and link the package
- A project directory that already has an MCP configuration
- An upstream MCP server definition you are authorized to use
Steps
- Run
mcp-slim-guard startin the project so the default Compact mode is active. - Make the tool call through the host and note the
result_refin the first response. - Call
read_resultwith aqueryto get up to three bounded local fragments. - Or call
read_resultwithoutqueryand follownext_cursorto recover the exact snapshot. Do not combinequerywithcursor.
Prompt for your agent
In the lennney/mcp-slim-guard setup at <project-path>, trigger a tool call that returns a large result and capture its result_ref. Use read_result with a query to retrieve matching fragments, and confirm the fragments come from the same snapshot without a second upstream invocation.
Caveat. Recovery only works on responses that need recoverable delivery, and the README does not describe the size threshold that triggers it.
Use this in your AI agent
Ask your agent
Using the Deduped MCP, get the setup steps for lennney/mcp-slim-guard and "Recover oversized MCP results without re-running calls"
Add Deduped to Claude Code once
claude mcp add --transport http deduped https://dedupedai.com/mcp
Every Monday: the 5 hottest repos and exactly how to use them. Free, one email a week, unsubscribe anytime.
Get them freeShare: X LinkedIn Hacker News Email Copy link
Third-party repos we don't own. AI-written use cases can be wrong. Disclaimer